Vercel just shipped Is Agentic, a free public tool that answers a question plenty of teams are quietly panicking about: can an AI agent actually use your website? The tool scores sites on four layers — discovery, access, usability, and payments — using a methodology built by Ora, an agent-experience research firm from era labs. Vercel handles the interface, storage, and scoring model, but the underlying checklist is Ora’s work.
The scoring isn’t arbitrary. Ora reverse-engineered 118 checks from real agent runs, not editorial opinion. Discovery carries 20 points across 15 checks, access 30 points across 41 checks, usability 40 points across 56 checks, and payments 10 points across 6 checks. Vercel then regroups those into its own model: Essential checks share an 80-point pool, Recommended checks 20 points, and Emerging signals add a bonus capped at five points that never penalizes when absent.
What’s genuinely smart here is the applicability logic. Recommended checks only activate when scan evidence positively identifies an API, OAuth flow, GraphQL endpoint, MCP server, developer portal, or commerce surface. A marketing site won’t get dinged for missing an interface it never claimed to offer. Each finding ships with observed evidence and a recommendation, and the ‘Prompt to fix’ feature converts those recommendations into an implementation brief for a coding agent.
There’s no API key, no billing account, no paid tier. You can hit the browser, run npx is-agentic <domain>, or wire the read-only API into CI as a regression gate. The report API lives at /api/v1/report with a 120-request-per-minute rate limit, and the MCP server exposes three tools for agents to pull reports and methodology. Vercel even practices what it scores: report pages render scores in initial HTML and serve a Markdown variant with proper Vary headers. For teams building agent-facing surfaces, this is the rare free tool that’s actually useful on day one.
